Clinical Infectious Diseases

MAJOR ARTICLE

Levofloxacin Prophylaxis During Induction Therapy for

Pediatric Acute Lymphoblastic Leukemia
Joshua Wolf,1,2,3 Li Tang,4 Patricia M. Flynn,1,2 Ching-Hon Pui,2,5,6 Aditya H. Gaur,1,2 Yilun Sun,4 Hiroto Inaba,2,5 Tracy Stewart,1 Randall T. Hayden,6
Hana Hakim,1 and Sima Jeha2,5
Departments of 1Infectious Diseases, 4Biostatistics, 5Oncology, and 6Pathology, St. Jude Children’s Research Hospital, and 2Department of Pediatrics, University of Tennessee Health Science
Center, Memphis; 3Department of Paediatrics, University of Melbourne, Parkville, Victoria, Australia

Background. Infection is the most important cause of treatment-related morbidity and mortality in pediatric patients treated
for acute lymphoblastic leukemia (ALL). Although routine in adults with leukemia, antibacterial prophylaxis is controversial in
pediatrics because of insufficient evidence for its efficacy or antibiotic choice and concerns about promoting antibiotic resistance
and Clostridium difficile infection.
Methods. This was a single-center, observational cohort study of patients with newly diagnosed ALL, comparing prospectively
collected infection-related outcomes in patients who received no prophylaxis, levofloxacin prophylaxis, or other prophylaxis during
induction therapy on the total XVI study. A propensity score–weighted logistic regression model was used to adjust for confounders.
Results. Of 344 included patients, 173 received no prophylaxis, 69 received levofloxacin prophylaxis, and 102 received other
prophylaxis regimens. Patients receiving prophylaxis had longer duration of neutropenia. Prophylaxis reduced the odds of febrile
neutropenia, likely bacterial infection, and bloodstream infection by ≥70%. Levofloxacin prophylaxis alone reduced these infections,
but it also reduced cephalosporin, aminoglycoside, and vancomycin exposure and reduced the odds of C. difficile infection by >95%.
No increase in breakthrough infections with antibiotic-resistant organisms was seen, but this cannot be excluded.
Conclusions. This is the largest study to date of antibacterial prophylaxis during induction therapy for pediatric ALL and the
first to include a broad-spectrum fluoroquinolone. Prophylaxis prevented febrile neutropenia and systemic infection. Levofloxacin
prophylaxis also minimized the use of treatment antibiotics and drastically reduced C. difficile infection. Although long-term anti-
biotic-resistance monitoring is needed, these data support using targeted prophylaxis with levofloxacin in children undergoing
induction chemotherapy for ALL.
Clinical Trials Registration. NCT00549848
Keywords. prophylaxis; leukemia; child; levofloxacin; Clostridium difficile.

Infections remain the most frequent cause of serious treat-
ment-related morbidity and mortality in children and adoles-
cents with acute lymphoblastic leukemia (ALL); currently, up to
4% of children with ALL die of infection [1–4]. Even nonfatal
infections can result in permanent end-organ damage, contrib-
ute to chemotherapy delay or modification, and increase anti-
biotic exposure [5]. Most serious infections occur during the
relatively short induction phase of chemotherapy [2].
Primary antibacterial prophylaxis during chemotherapy-re-
lated neutropenia in adults reduces clinically documented
infection, microbiologically documented infection, and infec-
tion-related mortality [6]. The US National Comprehensive
Cancer Network recommends antibacterial prophylaxis with
levofloxacin, a broad-spectrum fluoroquinolone, for adult
Received 11 May 2017; editorial decision 7 July 2017; accepted 24 July 2017; published online
September 14, 2017.
Correspondence: J. Wolf, Department of Infectious Diseases, St Jude Children’s Research
Hospital, 262 Danny Thomas Pl, MS 320, Memphis, TN 38105 (Joshua.Wolf@stjude.org).
Clinical Infectious Diseases®  2017;65(11):1790–8
© The Author 2017. Published by Oxford University Press for the Infectious Diseases Society
of America. All rights reserved. For permissions, e-mail: journals.permissions@oup.com.
DOI: 10.1093/cid/cix644
patients with acute leukemia. In addition, findings of animal
studies suggesting that colonization with Clostridium difficile
might be reduced by replacing β-lactam antibiotics with fluoro-
quinolones [7] are intriguing because hospital-acquired C. dif-
ficile infection is associated with a 6.7-fold increase in the odds
of mortality in hospitalized children [8]. However, no clinical
studies have shown that fluoroquinolone prophylaxis can pre-
vent C. difficile infection in high-risk patients.
In the pediatric ALL population, primary antibacterial
prophylaxis is controversial, because data supporting its efficacy
and safety are sparse. There are 2 published studies of antibac-
terial prophylaxis with fluoroquinolones for pediatric ALL. An
unadjusted retrospective analysis, from Saudi Arabia, reported
reduced bloodstream infection (BSI) and other infections with
ciprofloxacin prophylaxis during delayed intensification [9].
However, a randomized controlled trial of ciprofloxacin proph-
ylaxis during induction therapy in Indonesia found a marked
increase in the risk of fever, BSI, and death [10]. No published
studies to our knowledge conducted in the pediatric population
have evaluated the use of broad-spectrum fluoroquinolones
(which may have greater efficacy than narrow-spectrum agents

1790 • CID 2017:65 (1 December) • Wolf et al

such as ciprofloxacin), compared fluoroquinolones with other
prophylaxis regimens, or assessed the impact of antibacterial
prophylaxis on C. difficile infection. The current study aimed
to identify the effects of antibacterial prophylaxis during induc-
tion therapy for pediatric ALL on antibiotic exposure, infection,
and febrile neutropenia, and on C. difficile infection.
METHODS
This was an observational cohort study of patients receiving
induction therapy for newly diagnosed ALL. Infection-related
adverse event data were collected prospectively by the study team
and evaluated by cross-referencing with the electronic medical
record and institutional microbiology and pharmacy databases.
Data describing prophylaxis prescribing and antibiotic expo-
sure were collected retrospectively. The study was approved by
the St Jude Children’s Research Hospital Institutional Review
Board (XPD04-070).
Inclusion and Exclusion Criteria
Eligible for inclusion were all patients enrolled in the total XVI
study for newly diagnosed ALL (ClinicalTrials.gov identifier:
NCT00549848) at St Jude Children’s Research Hospital from the
opening of the trial (29 October 2007) who completed induc-
tion therapy before 6 January 2016. Similarly to a previous
study, total XV [11], induction therapy was administered over
42 days and comprised 4 weeks of prednisone, 4 doses of weekly
vincristine, 2 doses of weekly daunorubicin, and 1 or 2 doses
of PEG-asparaginase, followed by 2 weeks of cyclophospha-
mide, cytarabine, and thioguanine, plus intrathecal chemother-
apy according to risk category (Supplementary Table S1). The
induction phase could be modified or prolonged in response to
infection or other complications. Participants with clinically or
microbiologically documented infection before induction ther-
apy initiation or who had fever before induction therapy requir-
ing prolonged antibiotic therapy (>4 days) were excluded from
this analysis to avoid misclassification of antibacterial treatment
as prophylaxis. Similarly, patients who developed febrile neutro-
penia or suspected infection during the first 7 days of induction
or after <2 days of neutropenia were excluded because there was
insufficient opportunity to initiate primary prophylaxis.
Definition of Primary Prophylaxis
Primary prophylaxis was defined as the administration of sys-
temic antibacterial agents during induction therapy, before the
first episode of suspected or proven infection or febrile neutro-
penia, with the intent to prevent infection. Appropriate peri-
operative or Pneumocystis pneumonia prophylaxis and empiric
short courses of antibiotic given for fever at presentation or
for clinical events other than infection or febrile neutropenia,
such as vomiting, headache, or vertigo, were not considered
prophylaxis. From 2007 until July 2014, antibacterial prophy-
laxis was provided at the discretion of the treating clinician. It
typically consisted of cefepime, ciprofloxacin, or vancomycin
plus cefepime or ciprofloxacin and was initiated after the first
documented neutropenia after chemotherapy. From August
2014, institutional guidelines recommended levofloxacin
prophylaxis during all episodes of neutropenia expected to last
≥7 days.
Management of Febrile Neutropenia and Suspected Infection
The recommended febrile neutropenia treatment was cefepime
(or ceftazidime plus vancomycin if cefepime was unavailable),
with addition of vancomycin or an aminoglycoside for speci-
fied indications [12, 13] and substitution of meropenem for
cefepime in patients who had recently received cefepime or had
suspected intra-abdominal or Bacillus cereus infection [13].
These guidelines did not change during the study period.
Definitions Related to Infection
The Common Terminology Criteria for Adverse Events, ver-
sion 3.0 [14], were used to capture all infection-related adverse
events in the study database, and standard definitions were used
to further categorize episodes (Table 1). Antibiotic exposure
was quantified as “antibiotic days,” “specific antibiotic days,”
and “cumulative antibiotic exposure.” Antibiotic days were cal-
culated for each patient as the simple proportion of induction
days on which ≥1 systemic antibacterial antibiotic was admin-
istered (excluding Pneumocystis pneumonia prophylaxis), and
specific antibiotic days were calculated as the proportion of
induction days on which a specific antibiotic was administered
(cefepime and ceftazidime were combined). Cumulative antibi-
otic exposure was calculated as the sum of all specific antibiotic
days divided by the number of induction days, accounting for
the potential additive effect of coadministered antibiotics.
Statistical Methods
Fisher’s exact test was used to compare proportions, and
Kruskal-Wallis or Wilcoxon-Mann-Whitney tests were used to
compare medians. Associations between primary antibacterial
prophylaxis and clinical outcomes were assessed using multiple
logistic regression, and Firth’s penalized likelihood approach
was used to address the problem of complete separation [20].
Propensity to receive prophylaxis was estimated with a multi-
nomial logistic regression model including age, sex, race, Down
syndrome, and leukemia type as predictors. Then, a logistic
regression model was used to model prophylaxis groups and
duration of profound neutropenia as predictors, with inverse
probability weighting based on the propensity score. Kaplan-
Meier estimates of disease-free survival were used to graph the
time to first event for infection events in prophylaxis groups
and were compared using the log-rank test. Statistical analysis
was performed with SAS (version 9.4; SAS Institute), and differ-
ences were considered significant at P < .05 (2 tailed). No power
calculation was performed.
Prophylaxis in Pediatric ALL • CID 2017:65 (1 December) • 1791
Table 1. Definitions Related to Infections

Outcome Definition

Neutropenia Absolute neutrophil count ≤500/μL

Profound neutropenia
Febrile neutropenia
Microbiologically documented
infection
Clinically documented infection
Likely bacterial infection
Bloodstream infection
Severe sepsis
Urgent intervention
Clostridium difficile infection
Enterocolitis
Absolute neutrophil count ≤100/μL
Core body temperature ≥38.3°C or ≥38.0°C for ≥1 h in context of neutropenia [15, 16]
Diagnosed bacterial, viral, fungal, or parasitic infection, with supportive microbiological evidence, such as a positive culture,
antigen or PCR test results, or characteristic histopathological findings [15]
Infection diagnosed by the treating clinician for which a specific microbial cause could not be demonstrated [15]
Any infection with a microbiologically documented bacterial cause or that was clinically documented in categories typically
attributed to bacterial infection, including pneumonia, skin and soft-tissue infection, osteomyelitis or myositis, enterocolitis,
otitis media or externa, sinusitis, epididymo-orchitis, CVC pocket or tunnel infection, pharyngitis, perianal abscess or celluli-
tis, peritonitis, lymphadenitis, or culture-negative sepsis
Any infection caused by a recognized pathogen that was isolated from ≥1 blood culture in the context of a compatible clinical
illness; common commensal bacteria were included if identified from multiple culture sets, or if a single blood culture set
was collected before start of antibiotic therapy and the result deemed clinically significant by the treating clinician [17, 18]
Infection in conjunction with severe dysfunction of cardiovascular or respiratory systems or ≥2 other organ systems [19];
fever or elevated WBC count not required
Infection requiring admission to the intensive care unit, fluid resuscitation, or supplemental oxygen or associated with clini-
cian diagnosis of septic shock
Identification of C. difficile toxin in stool by enzyme immunoassay (from 2007 to April 2010) or toxin gene PCR (from April
2010 onward) in the presence of diarrhea
Syndrome of abdominal pain in conjunction with radiologically documented bowel-wall thickening or microbiologically docu-
mented C. difficile infection

Abbreviations: CVC, central venous catheter; PCR, polymerase chain reaction; WBC, white blood cell count.

RESULTS documented infection, enterocolitis, C. difficile infection, likely

Of 505 patients assessed for eligibility, 161 were excluded because
of the onset of suspected infection before induction therapy
(n = 64), within the first 7 days of induction (n = 41), or after <2
days of neutropenia (n = 19), or because of fever at presentation
requiring prolonged treatment (n = 37). Of the remaining 344
patients, 173 received no primary prophylaxis, 69 received levo-
floxacin prophylaxis, and 102 received other prophylaxis (Table
2 and Supplementary Figure S1). Other prophylaxis regimens
comprised cefepime, ciprofloxacin, or vancomycin plus cefepime
or ciprofloxacin (Supplementary Table S2). The median dura-
tions of neutropenia and profound neutropenia were longer in
patients receiving prophylaxis (Table 2), so duration of profound
neutropenia was included with propensity score in multivariate
analyses. Analyses related to C. difficile infection also included
duration of meropenem exposure, because this drug was tar-
geted by antimicrobial stewardship interventions.
Total antibiotic days and cumulative antibiotic exposure were
significantly increased in patients who received prophylaxis
(P < .001) (Figure 1). The pattern of antibiotic use in patients
receiving levofloxacin differed from that in patients receiving
no prophylaxis; levofloxacin exposure increased, but there was
a concomitant reduction in exposure to cefepime/ceftazidime,
vancomycin, meropenem (P < .001 for all comparisons), and
aminoglycosides (P = .002) (Figure 1).
Table 3 shows the incidence of infection-related adverse events
in this population. After adjustment for other covariates, patients
receiving any antibacterial prophylaxis were significantly less
likely than those receiving no prophylaxis to have febrile neu-
tropenia, clinically documented infection, microbiologically
bacterial infection, or BSI (Table 4) [12]. Reductions in C. difficile
infection and enterocolitis occurred only in patients receiving
fluoroquinolone prophylaxis; patients receiving cefepime-based
prophylaxis had a higher risk of both (Supplementary Figure S2
and Supplementary Table S3). Multiple logistic regression analy-
sis (Supplementary Table S4) and Kaplan-Meier analysis showed
similar results (Figure 2 and Supplementary Figure S3).
Levofloxacin prophylaxis alone was associated with a sim-
ilar, statistically significant reduction in febrile neutropenia,
microbiologically documented infection, and likely bacterial
infection. (Table 5) There was also significantly less enterocolitis
and C. difficile infection (Table 5). Multiple logistic regression
analysis (Supplementary Table S4) and Kaplan-Meier analysis
showed similar results (Figure 2 and Supplementary Figure S4).
Compared with patients receiving other prophylaxis reg-
imens, patients receiving levofloxacin prophylaxis had a simi-
lar risk of febrile neutropenia, clinically documented infection,
microbiologically documented infection, likely bacterial infec-
tion, and BSI (Table 5). However, levofloxacin prophylaxis was
associated with a significantly lower rate of C. difficile infec-
tion (adjusted odds ratio, 0.04; 95% confidence interval, <.01 to
.36; P < .001) (Table 5). There was some heterogeneity among
alternative regimens: Pooled ciprofloxacin-based regimens,
predominantly ciprofloxacin plus vancomycin, seemed to be
superior for some outcomes, but no individual regimen was
significantly superior to levofloxacin (Supplementary Figure S2
and Supplementary Table S3). Multiple logistic regression analy-
sis (Supplementary Table S4) and Kaplan-Meier analysis (Figure
2 and Supplementary Figure S4) again showed similar results.

1792 • CID 2017:65 (1 December) • Wolf et al

Table 2. Characteristics of Included Patients With or Without Prophylaxis

Patients, No. (%)a

Levofloxacin Prophylaxis
Characteristic No Prophylaxis (n = 173) (n = 69) Other Antibiotic Prophylaxis (n = 102) P Valueb

Age, median (IQR), y 5.8 (3 –11.9) 6.8 (3.9–11.1) 7 (3.6 –11.9) .67
Age group .75
≥10 y 50 (29) 18 (26) 32 (31)
<10 y 123 (71) 51 (74) 70 (69)
Sex .59
Male 103 (60) 43 (62) 56 (55)
Female 70 (40) 26 (38) 46 (45)
Race .86
White 134 (77) 56 (81) 80 (78)
Others 3 (23) 13 (19) 22 (22)
Down syndrome 4 (2) 1 (1) 2 (2) >.99
ALL type .57
T 29 (17) 15 (22) 21 (21)
B 144 (83) 54 (78) 81 (79)
ALL risk category .14
Low 89 (51) 37 (54) 51 (50)
Standard 81 (47) 28 (41) 43 (42)
High 3 (2) 4 (6) 8 (8)
Duration of neutropenia, 17 (11–24) 18 (12–23) 20 (17–25) .002
median (IQR), d
Duration of profound 6 (2–13) 7 (4–12) 11 (5–16) .001
neutropenia, median
(range), d

Abbreviations. ALL, acute lymphoblastic leukemia; IQR, interquartile range.

a
Data represent No. (%) of patients except where otherwise specified.
b
Fisher exact test was used for categorical and Kruskal-Wallis test for continuous variables.

There were 28 episodes of bacterial BSI in 28 patients dur-
ing the study period (Supplementary Table S5). Although BSI
was less common in patients receiving prophylaxis, the caus-
ative organisms were similar (Supplementary Table S5), and
the proportions of BSI episodes complicated by severe sepsis
or requirement for urgent intervention did not differ between
patients receiving prophylaxis and those receiving none (2 of 9
[22.2%] vs 4 of 19 [21.1%], respectively; P > .99). One patient
receiving levofloxacin prophylaxis developed bacteremia caused
by an extended-spectrum β-lactamase–producing Escherichia
coli strain that was resistant to levofloxacin and cefepime.
DISCUSSION
In this largest study to date, primary antibacterial prophylaxis
during induction therapy for ALL was associated with markedly
reduced rates of febrile neutropenia, clinical or microbiologi-
cally documented infection, likely bacterial infection, and BSI.
Although prophylaxis increased overall antibiotic exposure,
patients receiving levofloxacin prophylaxis had less exposure
to cefepime/ceftazidime, vancomycin, and aminoglycosides.
Unexpectedly, prophylaxis with levofloxacin drastically reduced
the risk of C. difficile infection and all-cause enterocolitis.
The observed reduction in infection and febrile neutropenia
extends data obtained from adult populations. A meta-analysis
of antibacterial prophylaxis in predominantly adult patients
showed that prophylaxis reduced febrile episodes, documented
infection, infection-related mortality, and all-cause mortal-
ity [6]. Few studies of fluoroquinolone prophylaxis in children
with leukemia have been published; most showed a reduc-
tion in infection-related adverse events [9, 21–24], but some
showed a paradoxical increase in these complications [10, 25].
Trimethoprim-sulfamethoxazole prophylaxis can also reduce
bacterial infection [26] but was routine in all groups in this study.
The current study provides new information by comparing
prophylaxis with levofloxacin, a broad-spectrum fluoroquinolone,
to alternative prophylaxis regimens and by excluding patients who
had no opportunity to receive prophylaxis, carefully controlling
for potential confounders and examining C. difficile infection.
Our study has several strengths related to the institutional setting
and study design. All patients were treated according to a single
protocol at a single institution and were provided with high-qual-
ity supportive care [2, 27, 28]. We used prespecified standard
definitions of febrile neutropenia, clinically or microbiologically
documented infection (which includes nonbacterial infections),
and BSI [15]. We added the category of likely bacterial infection,
comprising infections typically attributed to bacteria.
Patients with early infection or febrile neutropenia were
excluded from analysis because they might falsely raise the

Prophylaxis in Pediatric ALL • CID 2017:65 (1 December) • 1793

Figure 1. Antibiotic exposure during induction for each antibiotic prophylaxis group. Data are shown in a box plot; each box plot illustrates the upper and lower quartile
(box), median (line inside box), adjacent values (whiskers), and outliers (open circles). Antibiotic exposure and cumulative antibiotic exposure were significantly greater in
patients receiving levofloxacin or other prophylaxis than in those receiving no prophylaxis (P < .001 for all comparisons). However, patients receiving levofloxacin prophylaxis
had less exposure to cefepime/ceftazidime, vancomycin, meropenem, or aminoglycosides (data not shown) when compared with those receiving no prophylaxis (P < .001,
P < .001, P < .001, and P = .002, respectively) or other prophylaxis (P < . 001, P < .001, P < .001, and P = .04, respectively).

apparent rate of infection in the no-prophylaxis group. Similarly,
excluding patients who had an infection before receiving
chemotherapy or who received prolonged treatment for fever
at presentation ensured that only potentially preventable out-
comes were analyzed. Not all patients with fever at presentation
were excluded, because fever occurs in about 59% of patients
and is rarely related to infection [29]. Detailed chart review
ensured appropriate classification of prophylaxis regimens.
Although observational studies of this type can be subject to
selection bias, we aimed to ameliorate such bias by applying a
propensity score–weighted logistic regression model to account
for differences between groups.
The observed reduction in enterocolitis and C. difficile infec-
tion associated with levofloxacin prophylaxis is important
because these conditions can cause serious illness, malnutri-
tion, chemotherapy delay, or poor absorption of medications
for cancer treatment, all of which can reduce the chance of
cure [8, 30–33]. Furthermore, 2 studies found that C. difficile
infection in hospitalized children markedly increased the risk
of all-cause mortality [8, 34]. Patients receiving levofloxa-
cin prophylaxis were protected from this infection, despite an
incongruous increase in their total antibiotic exposure [30, 32].
A recent study in children with ALL found that exposure to
antipseudomonal cephalosporin and β-lactam antibiotics, espe-
cially cefepime, was the most important contributor to the risk
of C. difficile infection, but this study did not assess the effects
of quinolone exposure [30].
In an animal model, fluoroquinolone antibiotics did not
impair resistance to colonization with C. difficile, whereas
β-lactam antibiotics did [7]. Therefore, using fluoroquinolone
prophylaxis to reduce exposure to other antibiotics might pre-
vent C. difficile infection, but this is the first study to show such
a benefit. Previous studies of quinolone prophylaxis in adults
showed no such effect on C. difficile infection [6], and some
studies have even identified fluoroquinolone exposure as a risk
factor for C. difficile infection [35]. The difference in the cur-
rent study may be related to the marked reduction in the use
of broad-spectrum antibiotic therapy or to predominance of
quinolone-susceptible C. difficile [2, 35].
Other potentially confounding variables that affect the risk of
C. difficile infection in children with leukemia include shorter
hospital stay and granulocyte colony-stimulating factor admin-
istration [36], but there was no change in discharge guidelines
during the study period, and granulocyte colony-stimulating
factor was not used in this protocol. Importantly, the institu-
tional rate of C. difficile infection did not change during the

1794 • CID 2017:65 (1 December) • Wolf et al

Table 3. Unadjusted Incidence of Infection-Related Adverse Events by Prophylaxis Group

Patients, No. (%)

Prophylaxis Regimen

Outcome No Prophylaxis (n = 173) Any Prophylaxis (n = 171) Levofloxacin (n = 69) Other Antibiotic (n = 102)

Febrile neutropenia 106 (61.3) 74 (43.3) 28 (40.6) 46 (45.1)

Febrile neutropenia without 60 (34.7) 49 (28.7) 19 (27.5) 30 (29.4)
CDI or MDI
Any documented infectiona 97 (56.1) 71 (41.5) 27 (39.1) 44 (43.1)
CDI 52 (30.1) 37 (21.6) 13 (18.8) 24 (23.5)
MDI 69 (39.9) 43 (25.1) 18 (26.1) 25 (24.5)
BSI 19 (11) 9 (5.3) 4 (5.8) 5 (4.9)
Clostridium difficile infection 17 (9.8) 9 (5.3) 0 (0) 9 (8.8)
Any enterocolitis 25 (14.5) 15 (8.8) 3 (4.3) 12 (11.8)
Likely bacterial infection 64 (37) 33 (19.3) 11 (15.9) 22 (21.6)

Abbreviations: BSI, bloodstream infection; CDI, clinically documented infection; MDI, microbiologically documented infection.
a
“Any documented infection” included CDI or MDI, with or without neutropenia.

study period (Supplementary Figure S5). The effect of prophy-
laxis during induction on C. difficile infection in postinduction
chemotherapy cannot be determined from this study.
The study has some limitations. Grouping alternative
prophylaxis regimens provides a robust comparison group
but can mask differences between heterogeneous regimens
(Supplementary Figure S2 and Supplementary Table S3). The
study was not powered to detect a difference in outcomes
between different fluoroquinolones, and the reduction in C.
difficile infection was seen in both levofloxacin and ciproflox-
acin-based prophylaxis regimens (Supplementary Table S3).
Until 2014, antimicrobial prophylaxis was provided at the
discretion of the treating physician, so the potential for con-
founding is an important issue, especially in the context of a
single-center study, which can amplify the effects of confound-
ers. Clinicians may have preferentially prescribed prophylaxis
to patients at higher risk of infection, but this confounding by
indication would be expected to mask, rather than inflate, the
benefits of prophylaxis.
Other than neutropenia duration, risk factors for infection
did not differ significantly between groups (Table 2); known
variables were addressed by a propensity score–weighted mul-
tivariate logistic regression analysis (Tables 4 and 5). Another
potential confounder is the time period, because routine lev-
ofloxacin was introduced in 2014 and the other groups were
treated in 2008–2014. There is no evidence that this explains
the differences between groups, because there were no trends
in infection rates in study patients who received no proph-
ylaxis before 2014 (Supplementary Figure S6) and no sig-
nificant decrease in institutional rates of hospital-acquired
infection or C. difficile infection during the entire study period
(Supplementary Figure S5).
Antimicrobial stewardship interventions introduced dur-
ing the study period included prospective audit with feedback
for meropenem and linezolid use; however, no intervention
aimed at reducing the use of vancomycin, cefepime/ceftazi-
dime, or aminoglycosides. Meropenem exposure was included
as a covariate in the analysis of C. difficile infections to reduce

Table 4. Propensity Score–weighted Multivariate Logistic Regression Analysis for Effectiveness of Primary Prophylaxis

Outcome Crude OR (95% CI) P Value Adjusted OR (95% CI)a P Value

b
Febrile neutropenia 0.48 (.31–.74) <.001 0.23 (.14–.40) <.001b
b
Febrile neutropenia with CDI 0.44 (.22–.89) .02 0.30 (.14–.65) .002b
b
Febrile neutropenia with MDI 0.40 (.23–.70) .001 0.25 (.14–.48) <.001b
CDI 0.64 (.39–1.05) .08 0.54 (.32–.90) .02b
MDI 0.51 (.32–.80) .004b 0.40 (.24–.65) <.001b
BSI 0.45 (.20–1.03) .06 0.30 (.13–.73) .008b
c
Clostridium difficile infection 0.51 (.22–1.18) .11 0.38 (.16–.93) .04b
b
Likely bacterial infection 0.41 (.25–.66) <.001 0.26 (.15–.45) <.001b
Any enterocolitis 0.57 (.29–1.12) .10 0.44 (.21–.91) .03b

Abbreviations: BSI, bloodstream infection; CDI, clinically documented infection; CI, confidence interval; MDI, microbiologically documented infection; OR, odds ratio.
a
The propensity score of prophylaxis groups was estimated using a multinomial logistic regression model with age, sex, race, Down syndrome, and leukemia type as predictors; a logistic
regression model was then used to model prophylaxis groups and exposure to profound neutropenia during induction as predictors of clinical outcomes, with inverse probability weighting
using the propensity score.
b
Significant difference (prophylaxis versus no prophylaxis; P < .05).
c
Also adjusted for exposure to meropenem.

Prophylaxis in Pediatric ALL • CID 2017:65 (1 December) • 1795

Figure 2. Kaplan-Meier analysis of time to infectious complications during induction for each antibiotic prophylaxis group. Patients receiving levofloxacin prophylaxis dur-
ing induction therapy for pediatric acute lymphoblastic leukemia had a lower cumulative incidence of C. difficile infection (A) than those receiving no prophylaxis (P = .008) or
other prophylaxis regimens (P = .01) and a lower cumulative incidence of enterocolitis (B) than those receiving no prophylaxis (P = .03). Patients receiving any prophylaxis had
a lower cumulative incidence of febrile neutropenia (C) (P < .001) and likely bacterial infection (D) (P < .001), but there was no significant difference in the cumulative incidence
of febrile neutropenia or likely bacterial infection between patients receiving levofloxacin and those receiving other prophylaxis regimens (P = .52 and P = .36, respectively).

the risk of confounding by these stewardship interventions. In
2015, a 2-month nationwide shortage of cefepime necessitated
a temporary modification of institutional guidelines to include
ceftazidime plus vancomycin as first-line therapy for febrile
neutropenia. Accordingly, ceftazidime and cefepime exposure
were merged in the final analysis to avoid a spurious reduction
in cefepime exposure during the shortage. Despite these efforts
to address and ameliorate the effects of identified variables,
confounding of the study by unidentified variables remains
possible.
The prolonged neutropenia in patients receiving proph-
ylaxis might be related to the indication for prophylaxis
and was, therefore, included in the multivariate analysis.
However, a lower neutrophil nadir was seen in an Indonesian
study of children receiving ciprofloxacin prophylaxis, raising
the possibility of neutropenia as an adverse effect of fluoro-
quinolones [10]. A related consideration is that infection-re-
lated mortality during leukemia therapy is more common
in low-income countries and malnourished patients, so our
findings may not be generalizable to those settings [10].
The implementation of antibacterial prophylaxis has
been limited by concerns regarding possible adverse conse-
quences, especially the development of antibiotic resistance
[37]. Although antibiotic-resistant infections did not signifi-
cantly increase in this study, the sample size and time period
were inadequate to measure the medium- or long-term risk.
Studies of antibacterial resistance after levofloxacin proph-
ylaxis have yielded mixed results. A meta-analysis found
no significant increase in quinolone-resistant infection
(relative risk, 1.2; 95% confidence interval, .8–1.7), but the

1796 • CID 2017:65 (1 December) • Wolf et al

Table 5. Propensity Score–Weighted Multivariate Logistic Regression Analysis for Effectiveness of Levofloxacin Prophylaxis

Outcome Crude OR (95% CI) P Value Adjusted OR (95% CI)a P Value

Levofloxacin vs no prophylaxis
Febrile neutropenia 0.43 (.24–.76) .004a 0.28 (.15–.52) <.001b
b
Febrile neutropenia with CDI 0.33 (.11–.99) .048 0.25 (.09–.69) .007b
Febrile neutropenia with MDI 0.44 (.21–.93) .03b 0.34 (.17–.71) .004b
CDI 0.54 (.27–1.07) .08 0.48 (.26–.91) .02b
MDI 0.53 (.29–.99) .045b 0.49 (.28–.88) .02b
BSI 0.50 (.16–1.52) .22 0.42 (.15–1.16) .09
Clostridium difficile infectionc 0.06 (<.01 to .48) .003b 0.03 (<.01 to .24) <.001b
Likely bacterial infection 0.32 (.16–.66) .002b 0.24 (.12–.48) <.001b
b
Any enterocolitis 0.27 (.08–.92) .04 0.23 (.08–.67) .008b
Levofloxacin vs other prophylaxis
Febrile neutropenia 0.83 (.45–1.54) .56 1.17 (.64–2.14) .60
Febrile neutropenia with CDI 0.64 (.19–2.15) .47 0.74 (.25–2.19) .59
Febrile neutropenia with MDI 1.16 (.48–2.82) .74 1.68 (.74–3.84) .21
CDI 0.75 (.35–1.61) .47 0.85 (.44–1.65) .63
MDI 1.09 (.54–2.19) .82 1.41 (.76–2.63) .27
BSI 1.19 (.31–4.61) .80 1.85 (.54–6.35) .33
C. difficile infectionc 0.07 (<.01 to .57) .008b 0.04 (<.01 to .36) <.001b
Likely bacterial infection 0.69 (.31–1.53) .36 0.85 (.41–1.74) .65
Any enterocolitis 0.34 (.09–1.26) .11 0.38 (.12–1.14) .08

Abbreviations: BSI, bloodstream infection; CDI, clinically documented infection; CI, confidence interval; MDI, microbiologically documented infection; OR, odds ratio.
a
The propensity score of prophylaxis groups was estimated using a multinomial logistic regression model with age, sex, race, Down syndrome, and leukemia type as predictors; then a
logistic regression model was used to model prophylaxis groups and exposure to profound neutropenia during induction as predictors of clinical outcomes, with inverse probability weighting
using the propensity score.
b
Significant difference (P < .05).
c
Also adjusted for exposure to meropenem.

follow-up period in these studies was insufficient to address Supplementary Data

the question of long-term resistance [6]. Other groups have
reported increased resistance, including cross-resistance,
after extended use of quinolones [38, 39], although this did
not necessarily negate the benefit of prophylaxis [38]. In the
current study, exposure to antipseudomonal β-lactam anti-
biotics, aminoglycosides, and vancomycin was reduced. This
suggests that fluoroquinolone prophylaxis might shift antibi-
otic use away from agents less crucial for treating infection,
thereby balancing the overall increase in exposure. Large-
scale long-term investigations are needed to assess this pros-
pect [39].
Ours is the first study to examine the effects of primary anti-
bacterial prophylaxis with levofloxacin on serious infections
and antibiotic exposure in children undergoing induction ther-
apy for ALL. Although all prophylaxis regimens, including lev-
ofloxacin, reduced the risk of febrile neutropenia and systemic
infection, levofloxacin prophylaxis also shifted antibiotic use
away from agents typically used to treat infection and dramat-
ically reduced the risk of enterocolitis and C. difficile infection.
There was no documented increase in breakthrough infections
with resistant organisms, but this remains possible. These find-
ings support the targeted use of levofloxacin prophylaxis in chil-
dren and adolescents with ALL who are undergoing induction
therapy, with close long-term monitoring of antibiotic resist-
ance patterns.
Supplementary materials are available at Clinical Infectious Diseases online.
Consisting of data provided by the authors to benefit the reader, the posted
materials are not copyedited and are the sole responsibility of the authors,
so questions or comments should be addressed to the corresponding author.
Notes
Acknowledgments. We thank Keith A. Laycock, PhD, ELS, Elaine
Tuomanen, MD, Jose Ferrolino MD, and Marnie Dorsey for support in the
preparation of the manuscript.
Financial support. This work was supported by the National Institutes
of Health (grants CA21765, CA36401-26S1, and GM92666) and by ALSAC.
Potential conflicts of interest. All authors: No reported conflicts of
interest. All authors have submitted the ICMJE Form for Disclosure of
Potential Conflicts of Interest. Conflicts that the editors consider relevant to
the content of the manuscript have been disclosed.
References
1. Christensen MS, Heyman M, Mottonen M, et al. Treatment-related death in
childhood acute lymphoblastic leukaemia in the Nordic countries: 1992–2001. Br
J Haematol 2005; 131:50–8.
2. Inaba H, Pei D, Wolf J, et al. Infection-related complications during treatment for
childhood acute lymphoblastic leukemia. Ann Oncol 2017; 28:386–92.
3. O’Connor D, Bate J, Wade R, et al. Infection-related mortality in children with
acute lymphoblastic leukemia: an analysis of infectious deaths on UKALL2003.
Blood 2014; 124:1056–61.
4. Slats AM, Egeler RM, van der Does-van den Berg A, et al. Causes of death–other
than progressive leukemia–in childhood acute lymphoblastic (ALL) and myeloid
leukemia (AML): the Dutch Childhood Oncology Group experience. Leukemia
2005; 19:537–44.
5. Fiser RT, West NK, Bush AJ, Sillos EM, Schmidt JE, Tamburro RF. Outcome of
severe sepsis in pediatric oncology patients. Pediatr Crit Care Med 2005; 6:531–6.
6. Gafter-Gvili A, Fraser A, Paul M, et al. Antibiotic prophylaxis for bacterial
infections in afebrile neutropenic patients following chemotherapy. Cochrane
Database Syst Rev 2012; 1:CD004386.

Prophylaxis in Pediatric ALL • CID 2017:65 (1 December) • 1797

 7. Buffie CG, Bucci V, Stein RR, et al. Precision microbiome reconstitution restores
bile acid mediated resistance to Clostridium difficile. Nature 2015; 517:205–8.
8. Sammons JS, Localio R, Xiao R, Coffin SE, Zaoutis T. Clostridium difficile infec-
tion is associated with increased risk of death and prolonged hospitalization in
children. Clin Infect Dis 2013; 57:1–8.
9. Yousef AA, Fryer CJ, Chedid FD, Abbas AA, Felimban SK, Khattab TM. A pilot
study of prophylactic ciprofloxacin during delayed intensification in children
with acute lymphoblastic leukemia. Pediatr Blood Cancer 2004; 43:637–43.
10. Widjajanto PH, Sumadiono S, Cloos J, Purwanto I, Sutaryo S, Veerman AJ.
Randomized double blind trial of ciprofloxacin prophylaxis during induction
treatment in childhood acute lymphoblastic leukemia in the WK-ALL protocol
in Indonesia. J Blood Med 2013; 4:1–9.
11. Pui CH, Campana D, Pei D, et al. Treating childhood acute lymphoblastic leu-
kemia without cranial irradiation. N Engl J Med 2009; 360:2730–41.
12. Freifeld AG, Bow EJ, Sepkowitz KA, et al. Clinical practice guideline for the use
of antimicrobial agents in neutropenic patients with cancer: 2010 update by the
Infectious Diseases Society of America. Clin Infect Dis 2011; 52:e56-93.
13. Lehrnbecher T, Phillips R, Alexander S, et al; International Pediatric Fever and
Neutropenia Guideline Panel. Guideline for the management of fever and neutro-
penia in children with cancer and/or undergoing hematopoietic stem-cell trans-
plantation. J Clin Oncol 2012; 30:4427–38.
14. National Cancer Institute. Common terminology criteria for adverse events v3.0
(CTCAE). NCI, NIH, DHHS. August 9, 2006.
15. Haeusler GM, Phillips RS, Lehrnbecher T, Thursky KA, Sung L, Ammann RA. Core
outcomes and definitions for pediatric fever and neutropenia research: a consensus
statement from an international panel. Pediatr Blood Cancer 2015; 62:483–9.
16. Santolaya ME, Alvarez AM, Becker A, et al. Prospective, multicenter evaluation
of risk factors associated with invasive bacterial infection in children with cancer,
neutropenia, and fever. J Clin Oncol 2001; 19:3415–21.
17. Centers for Disease Control and Prevention. Bloodstream infection event (cen-
tral line-associated bloodstream infection and non-central line-associated
bloodstream infection). Available at: http://www.cdc.gov/nhsn/PDFs/pscManu-
al/4PSC_CLABScurrent.pdf. Accessed 26 April 2017.
18. Centers for Disease Control and Prevention. NHSN organism list (all organ-
isms, top organisms, common commensals, MBI organisms, and UTI bacteria).
Available at: http://www.cdc.gov/nhsn/XLS/master-organism-Com-Commen-
sals-Lists.xlsx. Accessed 26 April 2017.
19. Goldstein B, Giroir B, Randolph A; International Consensus Conference on
Pediatric Sepsis. International pediatric sepsis consensus conference: definitions
for sepsis and organ dysfunction in pediatrics. Pediatr Crit Care Med 2005; 6:2–8.
20. Heinze G, Schemper M. A solution to the problem of separation in logistic regres-
sion. Stat Med 2002; 21:2409–19.
21. Cruciani M, Concia E, Navarra A, et al. Prophylactic co-trimoxazole versus norfloxa-
cin in neutropenic children–perspective randomized study. Infection 1989; 17:65–9.
22. Laoprasopwattana K, Khwanna T, Suwankeeree P, Sujjanunt T, Tunyapanit W,
Chelae S. Ciprofloxacin reduces occurrence of fever in children with acute leu-
kemia who develop neutropenia during chemotherapy. Pediatr Infect Dis J 2013;
32:e94–8.
1798 • CID 2017:65 (1 December) • Wolf et al
23. Sulis ML, Blonquist TM, Athale UH, et al. Effectiveness of antibacterial prophy-
laxis during induction chemotherapy in children with acute lymphoblastic leu-
kemia. Blood 2015; 126: 249.
24. Yeh TC, Liu HC, Hou JY, et al. Severe infections in children with acute leukemia
undergoing intensive chemotherapy can successfully be prevented by ciprofloxa-
cin, voriconazole, or micafungin prophylaxis. Cancer 2014; 120:1255–62.
25. Felsenstein S, Orgel E, Rushing T, Fu C, Hoffman JA. Clinical and microbiologic
outcomes of quinolone prophylaxis in children with acute myeloid leukemia.
Pediatr Infect Dis J 2015; 34:e78–84.
26. Rungoe C, Malchau EL, Larsen LN, Schroeder H. Infections during induction
therapy for children with acute lymphoblastic leukemia. the role of sulfameth-
oxazole-trimethoprim (SMX-TMP) prophylaxis. Pediatr Blood Cancer 2010;
55:304–8.
27. McCullers JA, Williams BF, Wu S, et al. Healthcare-associated infections at a chil-
dren’s cancer hospital, 1983–2008. J Pediatric Infect Dis Soc 2012; 1:26–34.
28. Pui CH, Evans WE. A 50-year journey to cure childhood acute lymphoblastic
leukemia. Semin Hematol 2013; 50:185–96.
29. Khurana M, Lee B, Feusner JH. Fever at diagnosis of pediatric acute lympho-
blastic leukemia: are antibiotics really necessary? J Pediatr Hematol Oncol 2015;
37:498–501.
30. Fisher BT, Sammons JS, Li Y, et al. Variation in risk of hospital-onset Clostridium
difficile infection across β-lactam antibiotics in children with new-onset acute
lymphoblastic leukemia. J Pediatric Infect Dis Soc 2014; 3:329–35.
31. Sandora TJ, Fung M, Flaherty K, et al. Epidemiology and risk factors for
Clostridium difficile infection in children. Pediatr Infect Dis J 2011; 30:580–4.
32. Stevens V, Dumyati G, Fine LS, Fisher SG, van Wijngaarden E. Cumulative antibi-
otic exposures over time and the risk of Clostridium difficile infection. Clin Infect
Dis 2011; 53:42–8.
33. Tai E, Richardson LC, Townsend J, Howard E, Mcdonald LC. Clostridium difficile
infection among children with cancer. Pediatr Infect Dis J 2011; 30:610–2.
34. de Blank P, Zaoutis T, Fisher B, Troxel A, Kim J, Aplenc R. Trends in Clostridium
difficile infection and risk factors for hospital acquisition of Clostridium difficile
among children with cancer. J Pediatr 2013; 163:699–705 e1.
35. Dingle KE, Didelot X, Quan TP, et al; Modernising Medical Microbiology
Informatics Group. Effects of control interventions on Clostridium difficile infec-
tion in England: an observational study. Lancet Infect Dis 2017; 17:411–21.
36. Sung L, Aplenc R, Alonzo TA, Gerbing RB, Lehrnbecher T, Gamis AS.
Effectiveness of supportive care measures to reduce infections in pediatric AML:
a report from the Children’s Oncology Group. Blood 2013; 121:3573–7.
37. Haeusler GM, Slavin MA. Fluoroquinolone prophylaxis: worth the cost? Leuk
Lymphoma 2013; 54:677–8.
38. Kern WV, Klose K, Jellen-Ritter AS, et al. Fluoroquinolone resistance of
Escherichia coli at a cancer center: epidemiologic evolution and effects of discon-
tinuing prophylactic fluoroquinolone use in neutropenic patients with leukemia.
Eur J Clin Microbiol Infect Dis 2005; 24:111–8.
39. Macesic N, Morrissey CO, Cheng AC, Spencer A, Peleg AY. Changing microbial
epidemiology in hematopoietic stem cell transplant recipients: increasing resist-
ance over a 9-year period. Transpl Infect Dis 2014; 16:887–96.